When you're starting out, staying nimble is paramount. Infrastructure decisions such as your choice of database can be costly: moving between relational databases is hard enough, and if you find you're better served by an entirely different paradigm, you're inviting a lot of re-work.

Event sourcing is a great solution to this, storing just the individual events in a *write model* and allowing you to construct the *read model* as and when you need it. Among other things, it allows you to delay the decision of how to read your data. Be warned: many tools used to store those events use custom formats, their own APIs or are otherwise just as costly and lock-in inducing as introducing a database in the first place.

This talk will demonstrate how to use your application's own logs as the event source, pumped into a log-capturing tool such as Fluentd. Writing machine-readable JSON to our output, we will use our logs as our database, keeping our options open and saving valuable time in the process. We will show you how to use this event stream to maintain application state, and also to produce reports and custom analysis.